Lambda

Lambda is a cloud GPU platform for AI developers, researchers, and organizations that need infrastructure for training, fine-tuning, and inference on NVIDIA GPUs.

Key capabilities include:

• On-demand GPU instances (1 to 8 GPUs) available in minutes, supporting NVIDIA options such as B200, H100, A100, GH200, and H200, with automation via web UI, API, or CLI.
• 1-Click Clusters for distributed workloads, scaling from 16 to 2,000+ interconnected GPUs (including HGX B200, H200, and H100) for terms ranging from 1 week to 3 years.
• Superclusters for very large deployments, offering 4,000 to 165,000+ GPUs (including GB300 NVL72, HGX B300, and B200) on 3+ year, single-tenant contracts.
• Built-in observability with live, minute-by-minute insights, plus pre-installed ML tooling such as CUDA, Jupyter, PyTorch, TensorFlow, and Keras.

Pricing: Paid, with pay-as-you-go instances starting at $0.50/hr per GPU (examples include $4.99/hr for B200; older 8-GPU V100 instances listed at $4.40/hr), plus longer-term cluster and supercluster contracts.

What makes it notable: it combines minute-billed on-demand instances with options to scale into managed clusters and single-tenant superclusters, and it highlights no egress fees and access to newer NVIDIA hardware for large-scale model training and inference.